A review of analogue case studies relevant to large-scale underground hydrogen storage

Underground Hydrogen Storage (UHS) has gathered interest over the past decade as an efficient means of storing energy. Although a significant number of research and demonstration projects have sought to understand the associated technical challenges, it is yet to be achieved on commercial scales. We highlight case studies from town gas and blended hydrogen storage focusing on leakage pathways and hydrogen reactivity. Experience from helium storage serves as an analogue for the containment security of hydrogen as the two gases share physiochemical similarities, including small molecular size and high diffusivity. Natural gas storage case studies are also investigated, to highlight well integrity and safety challenges. Technical parameters identified as having adverse effects on storage containment security, efficiency, and hydrogen reactivity were then used to develop high-level and site-specific screening criteria. Thirty-two depleted offshore hydrocarbon reservoirs in the UK Continental Shelf (UKCS) are identified as potential storage formations based on the application of our high-level criteria. The screened fields reflect large hydrogen energy capacities, low cushion gas requirements, and proximity to offshore wind farms, thereby highlighting the widespread geographic availability and potential for efficient UHS in the UKCS. Following the initial screening, we propose that analysis of existing helium concentrations and investigation of local tectonic settings are key site-specific criteria for identifying containment security of depleted fields for stored hydrogen.


Introduction
Recent climate reports dictate the use of alternative fuels such as hydrogen to mitigate against the rising global temperature trend (IPCC, 2018;Ferrier et al., 2022;Maggio et al., 2019).The use of intermittent green energy powering electrolysers (Biggins and Brown, 2022;Shiva Kumar and Lim, 2022;Dong et al., 2023;Hazrat et al., 2022), as well as other evolving techniques, such as biological (Nowotny and Veziroglu, 2011), waste gasification (Santasnachok and Nakyai, 2022;Zhao et al., 2023) and photoelectrocatalysis (Pitchaimuthu et al., 2022), for hydrogen production is aligned with these drivers (Fan et al., 2021;Raman et al., 2022;Li et al., 2022;Sasanpour et al., 2021) as hydrogen use, either as a fuel or energy carrier, can be significantly less polluting than fossil fuels (Bionaz et al., 2022;Marocco et al., 2023;Lebrouhi et al., 2022).At the same time, utilisation and conversion of the existing natural gas transmission system offer an opportunity to substantially reduce the cost of this transition (Dodds and Demoullin, 2013;Jayanti, 2022).By 2030, hydrogen production is expected to Fig. 1.Schematic representation of the main UHS options.From left to right: Lined rock cavern storage within hard crystalline rock, depleted hydrocarbon reservoir storage in the pore space of sedimentary porous rock overlain by an impermeable rock layer, salt cavern storage within a ductile salt formation, and aquifer storage within the pore space of a brine-filled formation overlain by a caprock layer.
Caverns constructed in hard rock, lined with steel and concrete, are referred to as lined rock caverns (Johansson et al., 2018).Lined rock caverns are typically excavated in igneous and metamorphic rocks, where the inner liner provides the gas tightness and chemical resistance, and the outer concrete layer creates a smooth surface for its tight adherence (Patanwar et al., 2024).Compared to salt caverns, lined caverns can exceed pressures of 200 bar (Patanwar et al., 2024), and are regarded as a means of combining hydrogen storage with fossil-free manufacturing of crude stainless steel (Damasceno et al., 2023).

Storage in porous media
In contrast to artificial salt caverns, storage in depleted reservoirs and aquifers occurs within the innate porosity of the reservoir rock, with sealing efficiency provided by a relatively impermeable layer, referred to as the caprock (Zeng et al., 2023).In depleted gas reservoirs, roughly 10%-30% of the natural gas in place remains trapped within the reservoir porosity (Carden and Paterson, 1979;Bragg and Shallenberger, 1976;Ahmed, 2010), known as residual gas, whereas hydrogen losses due to resident brine entrapment are more significant, particularly within aquifer formations (Zivar et al., 2021).Due to the effects of pressure on intra-and intermolecular interactions, significantly higher gas volumes can be stored in deep porous formations, making them ideal for long-term hydrogen storage (Tarkowski, 2019b;Gholami, 2023).A drawback of these sites, however, is the increased probability of hydrogen consumption through biological and geochemical reactions within the host rock (Zivar et al., 2021).Studies evaluating infrastructural requirements for storage in deep porous formations suggest that repurposing and/or retrofitting existing natural gas infrastructure is an advantageous strategy for ensuring flexible, economic delivery of hydrogen to meet seasonal demands (Lord et al., 2014;Mouli-Castillo et al., 2021;Simbeck and Chang, 2002).

Pilot UHS projects
In accordance with NGS practice, the injection of cushion gas is required to pre-pressurise the storage formation and assist hydrogen recovery (Tarkowski et al., 2021).NGS has been successfully carried out in subsurface formations and reflects some similarities to UHS, in terms of cyclicity and containment (Carden and Paterson, 1979).However, the unique characteristics of hydrogen, combined with an incomplete understanding of its behaviour in deep geological formations (Hagemann et al., 2015), means that further research is required to ensure safe and efficient implementation (Hagemann et al., 2015).More specifically, the physio-chemical properties of hydrogen requires stringent monitoring of containment security (Zivar et al., 2021), diffusion losses (Pan et al., 2021;Li et al., 2022) and viscous fingering effects (Paterson, 1983), as well as chemical and microbiological reactions (Panfilov, 2010).In the past decade, numerous pilot projects have been deployed to assess the feasibility of UHS in depleted reservoirs, aquifers, salts, and lined rock caverns.Table 1 summarises the chosen storage formations and key objectives of the following UHS ventures: H2STORE (Pudlo et al., 2013), HyUNDER (Simon et al., 2015), Hy-INTEGER (Pudlo and Henkel, 2016), HyCHICO (Perez et al., 2016), Evaluation of the induced effects on the storage formation and the optimal technical parameters for efficient storage of H 2 , CH 4 , and compressed air (Kabuth et al., 2016;Bauer et al., 2017).
Early development of models and tools, for the evaluation of storage capacities, monitoring of storage effects, and operation planning (Kabuth et al., 2016).

2012-2015 Salt Cavern
Geotechnical evaluation of H 2 and compressed air storage in salt caverns and mapping of energy storage capacity of salt caverns in NW Germany (Donadei et al., 2015;Pollok et al., 2015a).
240 caverns in NW Germany were found to be fit for H 2 and compressed air storage (Pollok et al., 2015b).
HyPSTER 2021 -Ongoing Salt Cavern Combination of a 1 MW electrolyser and a salt cavern with pure H 2 storage of 100 cycles in order to assess the thermodynamic and geomechanical effects (Anon, 2021b).The next phase involves the commercialisation of this green H 2 (Anon, 2021a).
HYBRIT 2022 -Ongoing Lined Rock Cavern Pure H 2 injection into a crystalline rock lined with steel and concrete at 250 bar (Johansson et al., 2018;Vattenfall, 2022).The size of the lined cavern is 100 m 3 and it lies at a depth of 30 m.The successful operation will assist in up-scaling the quantity of H 2 that can be potentially stored in such formations (Anon, 2022c;HYBRIT, 2022).
Underground Sun Storage (Pichler, 2019), ANGUS+ (Kabuth et al., 2016), InSpEE (Donadei et al., 2015), HyPSTER (Anon, 2021a), and HYBRIT (Johansson et al., 2018).Overall, these projects have provided an early insight into the feasibility of hydrogen mixture storage in deep porous formations (Perez et al., 2016;Pichler, 2019), the large availability of cavern formations with good UHS potential (Pollok et al., 2015a;Crotogino, 2013), the challenges associated with accurately simulating hydrogen dissolution in saline aqueous solutions (De Lucia et al., 2015), potential formation and wellbore integrity issues at high temperature, pressure, and salinity conditions (Henkel et al., 2017), and early development of tools for site-specific evaluation of underground storage projects (Bauer et al., 2017).They also denoted that further research is required to fully understand all associated challenges of UHS projects.Thus, in this study, we review existing cases of hydrogen, helium, and natural gas storage in the subsurface, to highlight the associated challenges that can adversely affect such a project.These challenges are taken into account when creating site-selection screening criteria, which are subsequently applied to a number of hydrocarbon fields across the UK continental shelf in order to obtain an early insight into its UHS potential.

Background
Research on UHS has been conducted for more than 50 years.Walters (Walters, 1976) in 1976 first suggested the technical, environmental, and economic feasibility of UHS in porous media based on the experience from underground NGS.Carden and Paterson (Carden and Paterson, 1979) assessed the losses during UHS and suggested that dissolution and diffusion in resident brine, chemical and microbiological reactions, mixing with cushion gas, macroscopic and microscopic gas trapping would predominantly occur during the initial injection/withdrawal cycles.Conversely, they estimated that caprock leakage, pump losses, and capillary hysteresis result in cyclic losses accounting for approximately 1%-3% loss of stored hydrogen per storage cycle (Carden and Paterson, 1979).Paterson (Paterson, 1983) investigated the impact of fingering on operating losses and demonstrated that it also exacerbates diffusion and dissolution losses due to the increased contact area between hydrogen and brine.

Reactions
Hydrogen losses due to chemical, microbial, and geochemical reactions necessitate additional screening criteria for UHS site selection (Panfilov, 2010;Heinemann et al., 2021a;Bo et al., 2021;Hassanpouryouzband et al., 2022;Thaysen et al., 2021;Saeed et al., 2023).While sandstone formations are expected to prohibit abiotic chemical reactions in the presence of hydrogen (Hassanpouryouzband et al., 2022), calcite-rich lithologies appear to favour dissolution (Bensing et al., 2022;Al-Yaseri et al., 2022), so their presence within the caprock and storage formation requires further site-specific investigation to ensure containment security.Hydrogen dissolution in brine, as well as being a potential loss itself, affects the type and rate of these reactions (Toleukhanov et al., 2015).Hence, a thorough understanding of its behaviour within the storage formation is imperative (Li et al., 2018).Li et al. generated a solubility model with respect to pressure, temperature, ionic strength, molar fraction, activity coefficient of hydrogen, fugacity coefficient, Henry's constant, and Poynting factor that estimated solubility and fluid density at representative UHS conditions, with an error margin of 5%-15% (Li et al., 2018).

Thermophysical properties
Alanazi et al. evaluated both cubic and PC-SAFT equations of state (EoS) with regard to their accuracy for modelling thermophysical properties of hydrogen blends at 0.01-100 MPa and 92-367 , for hydrogen molar fractions between 0.001 and 0.9, and found that the non-cubic equations reflected better agreement with the experimental data (Alanazi et al., 2022).Hassanpouryouzband et al. similarly found that the GERG-2008 EoS modelled with good accuracy the thermophysical properties of hydrogen mixtures (methane, nitrogen, carbon dioxide) over 0.01 MPa, 200-500 K, and hydrogen molar fractions up to 90% (Hassanpouryouzband et al., 2020).

Diffusion & sealing efficiency
High diffusivity of hydrogen creates a significant challenge for mitigating operating losses (Simbeck, 2003;Bardelli et al., 2014) and hydrogen diffusion in water is reported to increase with pressure and temperature, and thus depth, (Muhammed et al., 2022), hence its consideration is critical for effective UHS.Hydrogen diffusivity across the water-filled clay caprock has been examined through molecular dynamics simulations (Ghasemi et al., 2022), finding that trends in diffusion coefficient compared to pore size differed with respect to the net charge of the clay surface; highlighting additional complexities.Hemme and van Berk (Hemme and Van Berk, 2018) performed geochemical numerical modelling to study a number of potential UHS risks including hydrogen diffusion along the water-filled caprock.Using a diffusion coefficient of 5.13×10 −9 m 2 s −1 for hydrogen and an illite-rich siliceous caprock of 5% porosity, the authors found that after 30 years most of the leaked hydrogen was confined within 4 m from the caprock accounting for an overall loss of 25% (Hemme and Van Berk, 2018).Xue et al. (2022) studied the influence of caprock macro-cracks on sealing efficiency through a numerical 2-phase flow dual porosity model, using methane as an analogue for hydrogen, considering gas adsorption-desorption in the matrix and gas-brine displacement in the fractures only (Xue et al., 2022).Studying the migration of methane into the caprock, the authors found that the invading gas caused adsorption expansion within the matrix and reduced fracture openings, resulting in higher sealing efficiencies (Xue et al., 2022).The diffusion rate of hydrogen in water-filled, clay-rich caprock samples was experimentally measured by Michelsen et al. (2022), using core holder apparatus comprising two chambers at the same pressure, filled with hydrogen and nitrogen respectively.The resulting diffusion coefficients for the smectite-rich and the smectite-poor samples were 8.0×10 −11 m 2 ∕s and 18×10 −11 m 2 ∕s (Michelsen et al., 2022) respectively, indicating the decelerating effect of clay minerals on diffusion, and thus, hydrogen leakage rate.

Wettability studies
In assessing the wettability effect on containment security, Ali et al. studied the wetting behaviour of mica samples, as representative caprock formations, by measuring advancing and receding hydrogenbrine contact angles over pressures 0.1-25 MPa, temperatures 308-343 K, and organic acid (stearic acid) concentrations 10 −9 -10 −2 mol/L (Ali et al., 2021).The authors found that mica samples tended to become more intermediate-wet with increasing pressure, decreasing temperature, and increasing organic acid concentrations, but still aquaphilic, and thus efficient in terms of their containment security (Ali et al., 2021).The hydrogen and cushion gas wettability of various clay and shale types has been examined as an important feature for storage containment security, and it was found that their hydrophilic wettability is favourable for effective hydrogen storage (Al-Yaseri et al., 2021;Al-Mukainah et al., 2022).Hydrogen-brine-reservoir rock wettability characteristics have also been examined using relative permeability (Rezaei et al., 2022;Hashemi et al., 2021) and contact angle measurements (Hashemi et al., 2022), consolidating the nonwetting state of hydrogen compared to brine and thus favourable wettability characteristics.Pan et al. (2023) determined that the relative permeability hysteresis resulted in a lower withdrawal factor, but higher withdrawal purity.The authors also found that hydrogen recovery factor reduces with hydrophilicity, while the purity level is unaffected (Pan et al., 2023).

Production characteristics & cushion gas
The displacement and trapping characteristics of hydrogen and brine in a sandstone sample were studied by Thaysen et al. (2022) through computed micro-tomography over a pressure range of 20-70 bar.The study found that pore fluid pressure exhibited no influence on the initial hydrogen saturation during drainage (Thaysen et al., 2022).Conversely, higher pressure resulted in lower hydrogen recovery during imbibition which decreased from 80% at 20 bar to 57% at 70 bar, suggesting that deeper formations are likely less favourable for hydrogen storage (Thaysen et al., 2022).Regarding the maximum operating pressure, Alessa et al. proposed a probabilistic method of its determination through the initial fluid pressures and evaluated uncertainty through Monte-Carlo simulations (Alessa et al., 2022).The role and favourable characteristics of the chosen cushion gas have been thoroughly examined through numerical simulation studies with regard to various properties such as density difference, viscosity contrast, cost, hydrogen recovery performance, and geological factors (Feldmann et al., 2016;Pfeiffer et al., 2017;Heinemann et al., 2021b;Lysyy et al., 2021;Wang et al., 2022;Kanaani et al., 2022).Lysyy et al. conducted a numerical simulation study on a depleted hydrocarbon reservoir to investigate the effects of working to cushion gas ratio and found that for 60%-80% hydrogen mixtures, high recovery efficiency with low gas mixing was attained (Lysyy et al., 2021).The influence of cushion gas type was investigated by Kanaani et al. and by Saeed and Jadhawar through compositional numerical modelling on a depleted oil reservoir and a North Sea aquifer, respectively, using methane, nitrogen, and carbon dioxide as cushion gas (Kanaani et al., 2022;Saeed and Jadhawar, 2024).The authors found a negative correlation between cushion gas molecular weight and recovery efficiency, such that methane reflected the best results (Kanaani et al., 2022;Saeed and Jadhawar, 2024).Regardless of the cushion gas type gas-gas and gas-liquid mixing were worse over the first cycles (Kanaani et al., 2022).The same trend was found by Feldmann et al. (2016) who suggested that dispersion alongside diffusion exacerbate gas mixing.Gravity overriding and viscous fingering phenomena were found to be less important for depleted gas reservoirs than aquifers (Feldmann et al., 2016).Similarly, a number of numerical studies have been conducted for establishing the optimal cycle duration for the UHS process (Luboń and Tarkowski, 2022;Foh et al., 1979;Ershadnia et al., 2022;Jadhawar and Saeed, 2023).

Helium and NG as analogues
From a field-wide standpoint, apart from the small number of small-scale hydrogen storage projects, practical experience and resulting understanding of hydrogen behaviour in the subsurface stems from 'town gas' storage; a mixture typically comprising hydrogen, methane, carbon dioxide, carbon monoxide and nitrogen in varying compositions (Liebscher et al., 2016).Case studies based on the direct implementation of town gas and hydrogen mixture storage within porous formations are showcased in Section 3, focusing predominantly on potential losses of hydrogen.The low molecular weight and size of hydrogen mean that the risk of its diffusive leakage across the caprock is increased compared to natural gas (Simbeck, 2003;Reitenbach et al., 2015).From a molecular size and diffusivity vantage, helium displays the most similarities to hydrogen (Dhorali and Reddy, 2012).Therefore, relevant experience from underground helium storage and helium injection experiments (HIEs) is highlighted in Section 4. Relevant wellintegrity case studies from NGS are investigated in Section 5, as the low molecular weight and flammability of hydrogen suggest that stringent regulations are required to reduce the risk of leakage (Hussain et al., 2022).Hence, natural gas case studies of leaky wells serve as useful analogues.In Section 6, site selection screening criteria are developed assisted from experience from the reviewed case studies, which are implemented on depleted and near-depleted offshore gas fields of the UK continental shelf.

Underground town gas and Hydrogen Mixture Storage Case Studies
As shown in Table 2, in which some technical characteristics of underground hydrogen storage projects of various compositions are illustrated, salt caverns have been hitherto employed for high-purity hydrogen storage.The good mechanical properties of salt caverns such as tightness and ductility, the inability of bacteria to withstand the high salinity environment (Kireeva and Berest, 2012), and the high recoverability factors have led to their use as storage formations for hydrogen mixtures since the 1960s.The high flexibility achievable in these formations is another reason for their widespread use (Crotogino, 2013).As some of these caverns have been operating for many decades, such as the Teesside Dome, their efficiency and stability for storing hydrogen volumes of up to 95% purity are proven.
Despite this, depleted gas reservoirs have significantly higher storage capacities than salt caverns, and for those cases where hydrogen mixtures or natural gas have been stored in these formations, good results were obtained (Table 2).Subsurface aquifers offer similarly large storage capacities.For example, approximately 200 million m 3 of hydrogen was stored at the Beynes aquifer (Hugout, 1997;Blondin, 1994).As hydrogen is lighter, more mobile, and more diffusive than all other gases (except for helium) special consideration is required for its containment security.Experience from underground town gas storage, in which working gas volumes have a high hydrogen content, can be used to demonstrate the feasibility of hydrogen storage within these sites since few major caprock leakages have been reported in the literature (Liebscher et al., 2016).

Town gas storage at the Lobodice Aquifer, Czech Republic
The geological setting of the Lobodice aquifer will be analysed in this section as an outlier from the pattern explained above.Despite the fact that the Lobodice aquifer remained operational for three decades since the 1960s, leakage of hydrogen has been identified and reported (Buzek et al., 1994).
The Lobodice aquifer lies in the middle of the Moravian Carpathian Foredeep (Kopal et al., 2016).The storage of town gas took place within the Lower Badenian clastics, in a horst structure on top of faulted and weathered crystalline units (Fig. A.4).More than 50 wells have been drilled to date; one-third of those for observational purposes (Buzek et al., 1994).The overlying formation comprises clays of significant thickness (gross thickness ∼200 m).Although the field was apparently operating efficiently in terms of gas storage, geophysical studies conducted prior to 1990 (Cahelová et al., 1990;Dvořáková et al., 2001) were not sufficient to fully depict its tectonic setting or establish the depth of the Lower Badenian clastic formation.This was only attained in 2009 by a 3D seismic survey (Wright et al., 2009).
The seismic survey showed that previously undetected fracture systems and faults run through the crystalline formation and disrupt the basal clastic formation (Wright et al., 2009).Moreover, synsedimentary faults were traced in the Lower Badenian clays on the rim of the elevation (Kopal et al., 2016) which were considered essential for the sealing efficacy of the formation.Buzek et al. (1994) studied the methane content increase in the formation using chemical and carbon isotope analyses and showed that hydrogen leakage due to diffusion took place along with microbial and geochemical reactions (Šmigáň et al., 1990) favoured by the low formation temperature below 45 • C (Šmigáň et al., 1990) resulting in overall hydrogen loss of 10%-20% (Panfilov, 2016;Buzek et al., 1994).The authors concluded that the caprock is tight for all gases (N 2 , CO, CO 2 , CH 4 ) except hydrogen.However, when the study was conducted (Buzek et al., 1994), the fractures across the formation remained undetected, and it is therefore arguable that hydrogen diffusion along the fractures contributed to the observed losses.
This case study thus highlights that along with caprock lithology, fracture mechanisms are similarly essential screening criteria for future UHS site selection.The existence of the thick clay layer above the aquifer was considered sufficient evidence of a gas-tight caprock, however, this proved not to be the case.

Town gas storage at the Ketzin Aquifer, Germany
Town gas was stored at the Ketzin Aquifer, in a sandstone formation (∼200 m depth) (Panfilov, 2016) at low ionic strength (total dissolved solids of roughly 50 g/L) and temperature (< 36 • C) conditions, over-lain by a 100 m mudstone caprock (Förster et al., 2006).Gas losses of up to 60%, as a result of chemical and microbial consumption, have been reported (Liebscher et al., 2016) with substantial H 2 S generation (Wagner and Ballerstedt, 2013).More specifically, a constant loss of CO and a corresponding gain in CO 2 was observed between the injection and withdrawal periods over the 16-year operation of the storage site.As illustrated in Fig. A.5 (Liebscher et al., 2016), fluctuating gas mixture compositions were also observed, resulting in net gains in hydrogen and methane concentrations (Liebscher et al., 2016).
In contrast to the Lobodice Aquifer, where the chemical and biological reactions were known, identification of reactions taking place was not possible (Liebscher et al., 2016), partly due to random fluctuations in methane and hydrogen ratios.However, the fact that gas composition concerns could be addressed by modifying the injection and recovery rates (Liebscher et al., 2016), suggests the synergistic effect of physio-rheological parameters or, similar to the Lobodice aquifer, leakage across undetected fractures.

Town gas storage at the Beynes Aquifer, France
Hydrogen-rich (50%-60%) town gas storage was performed in the Beynes sandstone aquifer between 1956 and 1972 (Jahanbakhsh et al., 2024).Minor hydrogen losses due to pyrite reduction were reported during the first years of its storage operation (Foh et al., 1979;Anon, 2017a).More significant losses were attributed to the reduction of sulfates present in the in-situ brine (Foh et al., 1979), resulting in hydrogen sulfide generation in high concentrations (40 ppm) requiring the desulphurisation of withdrawn gas (Foh et al., 1979).During site conversion, natural gas was injected to displace all resident town gas, leaving only 1% of the original town gas within the formation, resulting in minimal mixing (Foh et al., 1979,?).

Hychico Hydrogen Storage Project at the Diadema Wind Park and Hydrogen Plant
In 2010, Hychico set out to produce green hydrogen in the highpotential wind power region of Diadema in Patagonia, Argentina, and store that hydrogen underground in nearby depleted gas and oil reservoirs (sandstone reservoir, clay caprock) (Perez et al., 2016).Storage of natural gas in the Diadema field had been successfully performed 10 years before its conversion, yielding high deliverability rates over its 7-month injection and 5-month production cycles (Rodriguez and Morisseau, 2003).
The injection project started in 2015, comprising several phases with specific goals, to better evaluate the feasibility of the project.These included: (i) natural gas injection with traces of hydrogen to evaluate reservoir properties and the sealing efficiency of the clays (ii) 10% hydrogen injection to assess any changes in reservoir properties and gas composition and (iii) injection of natural gas at higher pressures to assess the tightness of the formation (Perez et al., 2016).The early results indicate that high recoverability of the injected hydrogen can be attained, although no information as to the caprock tightness has been reported.Furthermore, the physiochemical conditions of the formation were reported to favour hydrogen conversion to methane, driving the launch of the Methanation Project, as a means of storing energy underground, in 2018 (Perez et al., 2018).

Underground Sun Storage Project
The Underground Sun Storage project began in 2013 by RAG Austria AG, to evaluate the UHS potential of the depleted Lehen HP3 A gas reservoir (Pichler, 2019;Bauer, 2017).The caprock of the selected depleted reservoir is a thick, gas-tight clay (Bauer, 2017).
Upon the successful completion of lab studies, a mixture of 10% hydrogen and 90% natural gas was injected into a small sandstone (Upper Austrian molasse basin) reservoir (Bauer, 2017).A total volume of 1.22 million m 3 of the natural gas -hydrogen mixture was injected over three months (Bauer, 2017).A further three month shut-in period was imposed to investigate the pressure and temperature response of the formation, as an indicator of hydrogen reactions.
Results suggested the thick clay stratum above the depleted gas reservoir provided adequate sealing efficacy.However, an overall loss of 18% of the injected hydrogen was attributed to diffusion, dissolution, and chemical reactions (hydrogen reduction) (Pichler, 2019).The latter are believed to be minor, as no significant change in temperature was observed which would have indicated intense reactions (Pichler, 2019).The technical report noted that such losses will be significant during initial cycles, and their effect will ''dilute'' over subsequent phases (Bauer, 2017).During these initial cycles, concentration gradient was the steepest, and resident brine unsaturated from hydrogen.The partial pressure of hydrogen (10%) implies that faster and more significant losses might be expected for a pure hydrogen storage scenario (Bauer, 2017).Despite this, no hydrogen sulfide was generated, indicating a lack of hydrogen consumption by sulfate-reducing bacteria, although the latter were present in the reservoir (Pichler, 2019).While no significant gravitational segregation was experienced in the sandstone formation between hydrogen and methane, this is likely a result of small reservoir thickness (Bauer, 2017).Finally, a testing pipe comprising all the individual wellbore components was constantly exposed to the exposed mixture (Pichler, 2019).No damage, corrosion, or degradation was observed in any of the individual components (cement, steel, elastomers) by the utilised mixture (Bauer, 2017).

Insight from underground Helium Storage & Helium Injection experiments
Helium is an odourless, tasteless, extremely volatile and chemically inert noble gas (Epple and Lave, 1982).As an inflammable gas that is both a good conductor of electricity and transmitter of sound, helium has a wide variety of applications (Anon, 1978).While its inertness does not comply with hydrogen's reactivity, the two molecules are similar in terms of their molecular size and diffusivity.In fact, helium has a smaller atomic radius and kinetic diameter (Dhorali and Reddy, 2012).Kinetic diameter, in particular, is related to diffusivity (Kunze et al., 2022), and as result, the two gases are highly diffusive.Both gases are markedly supercritical at representative downhole storage conditions (Littlejohn, 1993) (helium is more supercritical than hydrogen at representative UHS conditions), but, while their viscosities and densities are similar (Petersen, 1970), helium is more viscous than hydrogen.Physiochemical properties for hydrogen, helium and typical cushion gases are presented in Table 3.It is therefore suggested that experience from underground helium storage and HIEs could provide useful insights for hydrogen storage, particularly in regard to gas mobility, diffusivity, and containment.

National Helium Reserve, USA
Helium occurs in natural gas reservoirs and streams (Rogers, 1921), and its separation from the latter is achieved based on the boiling temperature differences between the gases (Speight, 2019).The concentration of helium in natural gas reservoirs ranges from 0.01-7.00%(Grynia and Griffin, 2016).The first subsurface reservoir for helium storage was established in Amarillo, Texas, by the United States Bureau of Mines in the early 1960s (Epple and Lave, 1982).This has been the only large-scale underground storage for helium to date, although there has been considerable recent interest in such storage in other countries (Provornaya et al., 2022;McElroy et al., 2022;Rapatskaya et al., 2020).Site selection was based on the existence of a helium-rich natural gas field in the formation (Eilerts and Sumner, 1973), which not only facilitated the repurposing of existing infrastructure but most importantly meant the sealing efficiency of the caprock for the highly mobile and diffusive gas was already demonstrated.

Geology of cliffside field, Amarillo, Texas
The Cliffside Field comprises 5 domes.Helium storage takes place in the Bush Dome, an anticlinal structure of Brown Dolomite and Panhandle Limestone of 112 m maximum thickness and 21 m mean reservoir thickness (Eilerts and Sumner, 1973).The Bush Dome lies at a depth of 1,100 m.The caprock is impermeable anhydrite of the Panhandle formation with an average thickness of 122m (Eilerts and Sumner, 1973).Reservoir porosity ranges between 4 and 20%, and the average permeability is approximately 7 mD (Eilerts and Sumner, 1973).Immediately above the storage trap lies the shale-rich Red Cave formation which contains substantial natural gas reserves (Tade, 1967).The existence of gas reservoirs at a lesser depth than the storage formation was considered beneficial as any potential leak may be confined within the overlying reservoir (Golovachev and Starokon, 2020).The water present in the field is considered to be residual and lies at the bottom of the anticline.

Production data
Helium storage (initially of 70% purity and later on 95%) began in 1963.Nitrogen was the predominant constituent of the remaining mixture (Tade, 1967), with traces of methane and hydrogen (Golovachev and Starokon, 2020).After several decades of natural gas production, reservoir pressure was lowered from 56 to 47 bar (Eilerts and Sumner, 1973) corresponding to roughly 18% native gas recovery.It was thus determined that during helium injection into the reservoir, simultaneous native gas production must take place in order to attain a sufficient storage volume.So as not to compromise the caprock seal, the original reservoir pressure of 56 bar was not exceeded throughout the project.Thus, the injected Helium volume was roughly equal to the produced natural gas.Four injection wells, located at the crest, and 20 deeper recovery wells were originally employed (Eilerts and Sumner, 1973).For effective helium dispersion management across the field, a numerical simulation model was utilised (Eilerts and Sumner, 1973).

Interface monitoring
The evolution of the interface positions after 13 years since the onset of injection is illustrated in Fig. A.6.It can be seen that Well A-13 experiences significant propagation front hysteresis compared to the other two injection wells.This is because Well A-13 runs through a low permeable zone within the Brown Dolomite with an average permeability of 0.1 mD, whereas the other two injectors are located in regions of much higher permeability.As a result, four fingers were established after 13 years of injection, causing a relatively unstable displacement pattern.Since those crest injectors would be also used for the subsequent recovery of helium, this natural gas fingering would affect both efficiency and purity of the retrieval.By converting a deeper producing well into an injector, this problem was effectively addressed (Eilerts and Sumner, 1973).

Helium breakthrough
Within the first year, 40% of the natural gas recovery wells were invaded by helium, at an average 1.87% vol (Tade, 1967).In addition to the low viscosity and high mobility of helium as a likely cause, the fact that breakthrough times for some of these wells were expected later and high velocities were established at the leading front, suggest that diffusion of helium in the methane-rich native gas occurred.Moreover, the flow of helium through high-permeability ''thief-zone'' channels can add to this dispersive response.The fraction of helium in the producers increased gradually over the period; after 20 months of continuous injection and production, it rose from 2.0 to 4.5% (Tade, 1967).

Later stage of field's life
The Bush Dome underwent continuous injection and recovery cycles until 1996, when the Helium Privatisation Act demanded full recovery of all underground helium reserves, over the following 20 years (National Research Council, 2010).From then until August 2002, helium recovery predominated, leaving an estimated 32  of helium remaining (Weinstein et al., 2003).According to Weinstein et al. (2003), the main factors affecting recovery are native gas mixing, well interactions, depletion, and loss of reservoir energy.Each of these can be effectively mitigated through prudent well placement, proper adjustment of flow rates, based on the geological data of the field (i.e., potential faults acting as thief zones, low permeability zones etc.), and the inherent fluid properties.An additional factor might be found in effective well conversion.Dual porosity compositional numerical modelling, undertaken by the National Research Council (National Research Council, 2010), attempted to optimise helium recovery after performing a history matching against observed pressure data and helium fractions, and found that even though the injection of helium at the central and topmost region of the formation is logical, the same wells should not be used during recovery (Weinstein et al., 2003).By converting some natural gas producers into helium producers, and shutting in some injectors during recovery, gas-gas mixing and helium recovery were optimised (Weinstein et al., 2003).The site performed stable and efficient helium storage for several decades with no reported minor or major leaking.

Helium gas injection experiment at the Reskajeage Farm Test Site, England
This field-scale experiment was conducted in 1992, for the purpose of understanding the mechanisms of gas migration, for subsurface radioactive waste material storage.The experiment took place at a disused quarry in the Reskajeage Test Site, Cornwall (Lineham et al., 1996).

Description of the injection experiment
Helium was continuously injected at a depth of 54 m, through a single inclined well for a period of 9 days, initially displacing water from the formation.Overall, 63 sm 3 of helium were injected into the formation.Over the subsequent 12-month period, soil-gas surveying was undertaken to identify locations of helium leakage (Lineham et al., 1996), using multi-port SS probes (Ball and Nicholson, 1983), buried at depths of 0.5 m.

Geological setting of the Reskajeage Farm Test Site
Lithology at the target depth comprises mainly Upper Devonian sandstones and siltstones (Hancock and North, 1989).These rocks are significantly fractured, with the hydraulically conducting fractures exhibiting a maximum aperture of 0.22 m close to the interface (Bourke et al., 1991).Hydrogeological studies found the fracture zone at which the helium injection took place had an average aperture of 19μm.Geophysical surveys indicated the existence of an impermeable sandstone layer 5 m below the surface, acting as a barrier.

Results of the experiment
A short-period helium anomaly was detected at the surface was detected during the injection, attributed to the increased pressure gradient and subsequent up-flow of helium.As soon as injection was stopped and pressure dissipated in the target formation, helium uptake ceased.Three months later, a lasting and greater leak became apparent; a daily release of 2 × 10 −3 m 3 ∕day of helium was detected, 0.0032% of the total injected volume (Lineham et al., 1996).The delayed helium emergence, of constant flux, was correlated with helium displacing the watersaturated fractured rock and/or helium transport in solution (Lineham et al., 1996).It was presumed that tiny apertures of the low permeability sandstone layer, close to the surface, prohibited vertical migration of helium, due to the extremely high capillary entry pressures, forcing it instead to laterally migrate around the barrier (Lineham et al., 1996).Formation of a helium cushion, just beneath the barrier, could have driven the lateral spreading, and consequent slow but constant emergence of helium at the surface.

Helium injection experiment in Lac du Bonnet Batholith, Manitoba, Canada
In 1989, helium was injected into an inclined fractured rock at a depth of 40 m.Soil-gas surveys were used to monitor the surface helium fluxes (Gascoyne and Wuschke, 1997), so as to model gas migration patterns over fractured, water-saturated rocks.The experiment was run on the Underground Research Laboratory lease area.

Description of the injection experiment
Helium gas was injected, at a constant pressure of 4.9 bar for 11 days (Gascoyne and Wuschke, 1997), through borehole B34 (Fig. A.7), with increasing gas recovery rate over this period, indicating the displacement of water (Gascoyne and Wuschke, 1997).The occurrence of helium at the surface was monitored with soil-gas surveys over a 40-day period.

Geological setting of the Lac du Bonnet Batholith
The target formation is a highly fractured granitic batholith, cut by a major inclined thrust fault (thrust FZ2, Figure 4).Above the FZ2 thrust significant sub-vertical fracturing occurs, whereas the 15 m area below the surface comprises clays and silts, and is considered impermeable (Davison and Kozak, 1988).

Results of the experiment
Helium breakthrough at the surface occurred 48 h after the onset of injection, with an average concentration of 10 ppm for the tested soils.Significant peaks were observed within a 40 m radius of the injection borehole, whereas traces of helium were measured within a radius of 200 m.The high helium concentrations were attributed to gas flow through near-vertical fractures, while the low helium concentrations were a result of migration across the inclined fracture zone (Gascoyne and Wuschke, 1997).While a downhole solution gas transport mechanism and exsolution near the surface are expected, where a significant gas bubble is established (as in Underground Gas Storage sites), vertical migration of continuous gas phase occurs (Gascoyne and Wuschke, 1997).The buoyant gas travels faster through vertical fractures and voids as a smaller quantity of water has to be displaced, resulting in the faster and higher concentration anomaly within close proximity to the wellbore.However, the upwards migrating gas through the highly inclined fracture zone was delayed by resident water, despite the fact the fracture zone had a much higher permeability compared to the overburden layers.

Injection experiment in the Arbia Fault System of the Siena Graben, Italy
Two gas injection tests were performed (helium and CO 2 ) in 1993 into a single fault plane within a shallow aquifer (Ciotoli et al., 2005) of 20 m depth.Gas occurrence at the surface and its behaviour were monitored through gas-soil surveys, soil exhalation flux and analyses of groundwater.

Description of the injection experiment
Helium was injected into the major inclined fault at a constant pressure, greater than the sum of the hydrostatic and capillary pressures (150 kPa) but lower than the lithostatic pressure.The 180 mm borehole, with an inclination of 63 degrees, was drilled to a 27 m depth.The injection lasted 45 min, pumping a total of 0.36 m 3 of helium.Soil-gas surveying, using portable probes (0.5 m depth) and exhalation flux, was conducted for 40 days after the injection phase (Ciotoli et al., 2005).

Geological setting of the arbia fault system
The experiment site belongs to the Arbia Fault System, Sienna Graben, comprising clays, sandy clays and sands with gravels (Ciotoli et al., 2005).A major inclined fault runs through the aquifer, in addition to other minor faults (see Fig. A.8).The average fracture spacing is 70 μm.The injection formation is a sandy aquifer, overlain by a low-permeability clay layer.The water table lies 7 m below the surface.

Results of the experiment
Gas injection at a pressure higher than the sum of hydrostatic and capillary pressures (2.4 bar) was sufficient to establish two-phase flow, and simulate upwards migration of helium.Gas breakthrough occurred within 36 h, with an average convective helium velocity 0.18 mm/s.Increased helium concentrations were recorded over a confined area of 100-150 m 2 , indicating that convective flow across the fault was the dominant mechanism of helium migration in this experiment (Ciotoli et al., 2005).

Experience from underground NGS well integrity issues
Hydrogen leakage poses a serious safety hazard (Zivar et al., 2021) and many studies suggest that safety concerns affect public perception of hydrogen as a fuel (Stalker et al., 2022;Crotogino, 2013).Thus, when determining a framework for safe and reliable UHS, well integrity and hydrogen leakage across the wellbore are important factors.Experience from well integrity failures, from underground NGS case studies, is explored in the following sections.UHS requires more stringent regulations with regards to well integrity compared to NGS due to potential microbial corrosion of the casing (Ugarte and Salehi, 2021) and embrittlement issues (Ugarte and Salehi, 2021; Ghosh et al., 2018).Most importantly, the high flammability of hydrogen-oxygen mixtures (Das, 2016) imposes additional safety considerations.The multi-decade experience from NGS can be deemed as a useful analogue for UHS, when considering well integrity.

The Stenlille Aquifer case in Denmark
The underground NGS facility of the Stenlille Aquifer was established in 1989 (Laier and Obro, 2009).Storage takes place within a compartmentalised anticlinal structure at a depth of 1500 m, comprising mainly sandstone (Gassum Formation) and mudstone, overlain by a 300 m claystone caprock (Laier, 2012).A total of 20 wells have been drilled: 14 for injection/withdrawal and 6 for observation, the latter in the periphery of the aquifer.The stored natural gas is methane-rich (91%) (Laier, 1989).
The heterogeneity of the aquifer formation and its large storage capacity (3 billion m 3 ) led to an increase in the number of wells drilled over the first years of storage site development (Laier and Obro, 2009).In August 1995, gas leakage occurred across a freshly drilled injection well, which was attributed to gas seeping into the annulus, displacing water and penetrating the cement and confining rocks (at 780 m depth) through small leaks over the casing (Laier and Obro, 2009;Laier, 2012).Over subsequent weeks, gas migrated to the surface, and small bubbles were observed in wellhead pit.Chemical analysis verified that the leaked gas had a comparable composition to that injected one.The absence of annulus pressure monitoring during injection, rendered the leak unidentifiable until observed at the surface.As a result, more stringent regulations for new wells were imposed (Laier and Obro, 2009).

The Aliso Canyon Leak Case in California, US
A number of NGS leaks have been attributed to well integrity issues (Michanowicz et al., 2017).The most prominent among them was a methane leak of ∼100,000 tonnes at the Aliso Canyon NGS facility, California, in 2016 (Conley et al., 2016).The Aliso Canyon Field was first developed as an oil reservoir during the 1930s and was converted into a gas storage field in the 1970s (Anon, 2016).The sedimentary storage formation lies at a depth of 2500 m and shales constitute the caprock (Anon, 2012(Anon, , 2019a)).All wells penetrate the Santa Susana and Honor Rancho faults, which are tectonically active, resulting in significant earthquakes of magnitude ∼6.7 in 1971 and 1994 (Anon, 2012).
The Aliso Canyon leak lasted 4 months and resulted massive residential displacement, as well as significant greenhouse gas emissions (Davis, 2018), and the well casing was identified as the cause (Michanowicz et al., 2017).The well responsible was drilled during the 1950s as an oil-producer and, although it was repurposed for NGS during the 1970s, it was susceptible to single point failure (Michanowicz et al., 2017) for two main reasons: its single casing was exposed to the adjacent rock formations over a depth of 2,000 m, facilitating gas migration (Lindeberg et al., 2017), and gas was both injected and recovered through the inner production tubing and outer well casing.8 well-kill attempts from the surface were unsuccessful, and the leak was eventually stopped by a relief well that successfully pumped heavy fluid into the leaky well (Anon, 2012).

Montebello Field Leak, California, US
The Montebello Field is a multi-zone compartmentalised oil and gas reservoir, discovered in the 1910s (Stolz, 1941).The reservoir rock chiefly comprises interbedded sands and shales, and the overlying caprock is a 600 m sandy shale (Stolz, 1941).During the 1960s, gas storage was attained in 3 reservoir zones, at depths between 1530 and 2280m (Miyazaki, 2009).During the 1970s, increased soil-gas concentrations were observed over a confined residential area (Miyazaki, 2009).Several extraction wells were drilled and, with additional remedial operations, the leak was mitigated (Miyazaki, 2009).The seepage site was adjacent to two abandoned wells (Miyazaki, 2009), within a larger area containing hundreds of abandoned wells (Chilingarian et al., 2003).The migration path of the stored gas was not accurately identified, but the abandoned, obsolete wells were considered a likely cause (Miyazaki, 2009).

Applying case studies to UHS
Findings from large-scale subsurface Town Gas and hydrogen-mixture storage projects suggest the feasibility of UHS in porous media, consolidated by the Hychico and Underground Sun demonstrations that initially indicate high hydrogen recovery, for hydrogen mixtures up to 10% (Perez et al., 2016;Bauer, 2017).However, both these and the geological helium storage experiments investigated within this review, highlight some of the key challenges for successful containment of hydrogen.
Understanding the geological controls required to ensure the efficient subsurface storage of hydrogen, such as caprock lithology and geothermal gradient, is essential for effective site selection and mitigating risks associated with containment.Evidence from analogous case studies investigated within this review offers valuable insights regarding the behaviour and flow mechanisms of hydrogen within porous media.Compounding these recommendations, we propose new high-level and low-level screening criteria for UHS sites.The former can be directly applied to reservoir databases, indicating the prospectivity of a particular region or field.We demonstrate such application to 96 depleted or near-depleted offshore gas fields on the UKCS (Fig. 2), using only public data.In contrast, the Low-level screening criteria derived from analogue case studies, as indicators of high containment security for hydrogen within the storage formation, can be only applied on a site-specific basis.

Caprock integrity
It is clear that sealing effectiveness of the reservoir caprock is of paramount importance for successful hydrogen containment.While this in itself is not a novel finding, evidence from helium storage case studies has provided new insights into potential migration and leakage pathways, and subsequently highlighted important considerations when selecting a UHS site.Potential leakage mechanisms comprise breakthrough into pore spaces once capillary pressures have been exceeded, diffusion of dissolved gas through saturated conduits within the caprock, migration through faults or fractures and finally, well leakage (Song and Zhang, 2013).
The complexity of sedimentary geology, particularly that found in fine-grained and heterogenous caprocks makes it challenging to determine a definite criterion for evaluating caprock integrity and tightness.The sealing ability of argillaceous caprocks, for example, depends on properties including permeability, capillary entry pressure and sorption capacity; however, as demonstrated in the Reskajeage experiment (Lineham et al., 1996), presence of a low-permeability overburden is not a guarantee of sealing efficacy, and thus cannot be used in isolation as proof of containment for hydrogen.Contrastingly, evaporites (salts and anhydrites) form tight traps but are vulnerable to brittleness, and subsequent fracturing, at shallower depths and lower temperatures (Song and Zhang, 2013).However, unless the formation is completely homogeneous, its deformation behaviour is difficult to accurately predict based on depth alone; hence, similarly, it cannot in isolation guarantee effective sealing for UHS.On a site-specific basis, focused analyses of tectonic setting, geomorphology, heterogeneities and micro-/fracturing patterns/orientations will better indicate sealing efficiency, together with historical evidence of successful helium containment within the reservoir (Danabalan et al., 2022;Halford et al.,  2023), Saboorian-Jooybari ( 2016), Zeyghami and Taghizadeh (2023), Lineham et al. (1996), Gascoyne and Wuschke (1997), Ciotoli et al. (2005), Danabalan et al. (2022), Halford et al. (2022), Hooker et al. (1985), Misra et al. (1988).
2022; Hooker et al., 1985;Ballentine et al., 1996).Experience from the National Helium Reserve also suggests that knowledge of local faults and low-permeability zones is crucial for mitigating physical dispersion of hydrogen (Tade, 1967).
The literature suggests that the sealing capacity of a caprock depends on dynamic capillary entry pressures, interfacial tensions and hydrogen-water-rock wetting characteristics (Ali et al., 2021;Esfandyari et al., 2022;Hosseini et al., 2022); all of which are influenced by the geochemical composition of the storage formation.Caprock lithology was therefore determined as a good initial screening criterion for UHS; consolidated by Lewandowska-Śmierzchalska et al. (2018), who proposed that caprock lithology is the most important geological site selection criterion.The study also suggested lithologies be ranked evaporitic, argillaceous and medium-permeability (sandstones, limestones, and dolomites), in order of reducing favourability.In addition to this, reservoirs with evaporitic caprocks but temperatures less than 100 • C, the determined brittle-ductile transition for evaporites (Zhuo et al., 2014), are considered less favourable due to their increased likelihood of fracturing, and thus compromised caprock integrity, caused by geomechanical processes and injection-induced pressure.

Diffusion
Diffusion, both as a dissolved gas through the caprock and through cushion gas-mixing, will compromise containment security and purity of stored hydrogen volumes, respectively.While some hydrogen migration into the caprock is inevitable, the diffusive losses are suggested to occur mainly within the first cycles (Carden and Paterson, 1979;Bauer, 2017).Moreover, as demonstrated in the Lac du Bonnet experiment, the presence of water-saturated strata overlying the reservoir constrains the upward migration of buoyant hydrogen (Gascoyne and Wuschke, 1997).In many depleted natural gas reservoirs, the caprock brine is already saturated with methane, which further inhibits hydrogen penetration (Amirthan and Perera, 2023).
The mechanisms of working gas-cushion gas mixing and viscous fingering are controlled by reservoir characteristics, including compressibility, porosity and permeability (vertical and lateral) (Sadeghi and Sedaee, 2022).In particular, the extent of mixing, which is considered to mainly affect the initial cycles (Carden and Paterson, 1979), also depends on the type and composition of the cushion gas, and can thus be excluded from high-level site selection.Furthermore, as in the case of the helium Cliffside storage site, gas mixing can be mitigated via effective well placement, adjustment of flow rates, and well conversion (Weinstein et al., 2003).Additionally, the Beynes aquifer case reported a mixing of 1% between remaining town gas and injected natural gas (Foh et al., 1979).At any rate, as a factor that can potentially account for losses up to 10% of the injected volume per cycle, over prolonged production, it should be factored into anticipated storage losses (Kanaani et al., 2022).
As a ubiquitous operational loss, yet of smaller magnitude compared to caprock integrity, well leakage, and biogeochemical reactions, diffusion is not considered an important high-level site-selection criterion.However, following an initial screening of prospective sites, further investigation should be undertaken to analyse the extent of hydrogen diffusion into caprock brines and determine an appropriate cushion gas to compliment the chosen lithostratigraphy.

Well integrity
Leakage through poorly sealed or disused wells also poses a significant risk to the successful containment of stored hydrogen and is therefore a key consideration when evaluating the effectiveness of a reservoir top seal.Each well constitutes an additional perforation through the caprock and is thus a potential path for vertical hydrogen migration.In terms of the integrity of plugged wells, the age of production operations is significant (Anon, 2017b); whether field reuse was considered at the time of well decommissioning, could indicate if wells were plugged effectively or if they are likely to compromise caprock integrity (Peecock et al., 2023).In both Aliso Canyon and Montebello natural gas leak cases, the reuse of abandoned wells was assumed responsible for the detected leakage (Michanowicz et al., 2017;Lindeberg et al., 2017;Miyazaki, 2009).The number and age of wells are therefore crucial site selection criteria; the fewer and younger, the more favourable for future UHS.Therefore, fields with more than 50 wells or whose operation began more than 50 years ago, have been excluded.Although some evidence exists on high pressure, temperature, and salinity conditions affecting well components (Henkel et al., 2017), no such criterion was imposed.Firstly, field-scale application of hydrogen storage, such as the Underground Sun Storage project resulted in no degradation or corrosion over these components (Pichler, 2019;Bauer, 2017), while other experimental studies focused on hydrogen effects on wellbore components at HPHT conditions found no obvious effect on cement and steel components (Boersheim et al., 2019;Iorio et al., 2022).

Microbial reactions
From town gas storage case studies, the significance of microbial impacts on UHS is clear; artificially elevating hydrogen concentrations within the subsurface could have adverse consequences for storage operations, including loss of hydrogen volume as high as 60% (Liebscher et al., 2016), reduced injectivity and recovery through diminished permeability and corrosion of infrastructure (Gregory et al., 2019).Estimating the risk of microbial growth within UHS reservoirs is thus crucial when selecting a suitable storage site, and can be achieved by understanding controls on hydrogen-consuming processes that occur within the subsurface.While nutrient availability will be ascertained on a case-by-case basis, indicators such as reservoir temperature, pressure, and salinity can be screened to eliminate those with optimal environmental conditions for microbial growth.Temperature greater than 60 • C was found to be inhibiting for the methanogenic consumption of hydrogen in the Lobodice town gas case (Šmigáň et al., 1990), whereas the low formation temperature (< 36 • C) accompanied by low salinity conditions (equivalent to 0.8 M NaCl) in the Ketzin aquifer (Förster et al., 2006) favoured the observed biochemical reactions.The same can be inferred for the Lobodice aquifer storage exhibiting similarly low temperatures between 25 • C and 45 • C (Wagner and Ballerstedt, 2013).Thaysen et al. (2021) found reservoir temperatures higher than 122 • C, salinities greater than 3 M (NaCl) and pH above 10.2 inhibit hydrogen-consuming microorganisms, whereas an upper-pressure limit to microbial life could not be determined within realistic reservoir conditions.However, as these parameters are not mutually exclusive, they concluded that sites with temperatures and salinities exceeding 55 • C and 1.7 M, respectively, are favourable for minimising microbial hydrogen losses (Thaysen et al., 2021).

Additional criteria
The site selection criteria for UHS, developed in this study, are applied to a database of 96 UK offshore gas fields with available data, together with the following additional parameters: Reservoir depths greater than 1000 m -subsurface depth and pressure increase, resulting in greater hydrogen density and thus efficiency of storage (Thiyagarajan et al., 2022).
Hydrogen storage capacity less than 120 TWh -high cushion gas and hydrogen injection requirements suggest that fields of such high potential hydrogen storage capacities will not prove economic for conversion to large-scale UHS facilities (Peecock et al., 2023).The higher heating value ( ) of hydrogen (39.4 kWh/kg (Engineering ToolBox, 2003b)) was used in the calculation of the energy storage capacity (), as shown in Eq. ( 1): where   and   are the density values of hydrogen and methane, respectively, at the initial reservoir and temperature conditions,    is the density of methane at standard conditions,  corresponds to the original gas in place volume, and   is the working gas ratio, defined later in Eq. ( 1) in Section 6.2.3.As the exact composition of most of the screened reservoirs is unknown, methane was used as a proxy for natural gas in the calculations, so that the storage capacities may slightly vary in reality.Threshold values for formation porosity and permeability of 2% and 0.1 mD, respectively, are typically used for gas reservoirs (Saboorian-Jooybari, 2016;Zeyghami and Taghizadeh, 2023).Since all 96 reservoirs successfully operated or still operate as gas production fields, they all naturally comply with these criteria.Moreover, the similarly highly mobile helium was successfully stored and withdrawn for decades in the Cliffside field, which has an average permeability of 7 mD and porosity values as low as 4% in parts of the field (Eilerts and Sumner, 1973), implying that less strict threshold values might be applied for hydrogen compared to natural gas.Similarly, in the Arbia Fault System HEI the clay layer of 0.003-0.007 permeability was conductive to helium flow (Ciotoli et al., 2005).

Tectonic setting
While caprock lithology and thickness, accompanied by ultra-low permeability values are good indicators of sealing efficiency it was demonstrated in the Reskajeage experiment (Lineham et al., 1996) that the presence of a low-permeability overburden is not a guarantee of sealing efficacy for highly diffusive gases such as helium and hydrogen, and thus cannot be used in isolation as proof of containment.In all the HEIs studied, helium migrated through fractures and microfractures displacing resident waters, especially during the helium stage, with breakthrough occurring faster through lower-permeability vertical faults than higher-permeability inclined faults (Lineham et al., 1996;Gascoyne and Wuschke, 1997;Ciotoli et al., 2005).The fact that helium diffusion markedly decreased after helium injection ceased (Gascoyne and Wuschke, 1997) suggests that reducing the target pressure during hydrogen unloading can reduce the migration rate across water-filled faults.Hence, depending on the existence of faults and fractures, the strategy of not exceeding the original formation pressure as employed in the Cliffside storage field (Eilerts and Sumner, 1973) may not be sufficient, and lower pressures should instead be applied.We thus suggest that understanding the tectonic setting, along with micro-/fracturing patterns/orientations of a prospective UHS formation, is essential for ascertaining sealing efficiency.This is particularly crucial where fluids have been injected to assist production and where new wells are to be drilled alongside existing ones.Further drilling may cause caprock fracking (Opedal et al., 2018), which may compromise containment security.A detailed 3D seismic survey of the Lobodice storage site exposed the existence of fractures that may have caused significant hydrogen leakage (Panfilov, 2016;Buzek et al., 1994).More than 50 wells were drilled within the site's lifetime (Buzek et al., 1994), which could have been a contributing factor to the extensive fracturing, although it has not been verified.

Native Helium
As a more diffusive gas than hydrogen, with a similarly high risk of leakage due to its small molecular size (Dhorali and Reddy, 2012;Halford et al., 2022), helium provides an excellent analogue for understanding the impact of hydrogen diffusion within porous reservoirs.The natural presence of helium within a hydrocarbon reservoir, as in the Cliffside storage site (Eilerts and Sumner, 1973), therefore provides a good indication that the caprock could be sealing for hydrogen also.Compared to reservoirs containing less diffusive compounds and larger molecules, those with historically increased helium concentrations demonstrate containment security to a greater extent unless there is ongoing helium migration into the trap at a rate exceeding leakage through the top seal (Danabalan et al., 2022).Literature suggests that assessing the tectonic and stratigraphic setting of the heliumbearing formation, coupled with gas sampling and subsequent analysis including noble gas and isotope analyses, can be effectively utilised to identify the migration path of helium and whether there is ongoing migration of helium or not (Danabalan et al., 2022;Halford et al., 2022;Hooker et al., 1985).In the UHS context, this can be used to verify whether there is an outflux of native helium and ascertain whether there is sufficient sealing efficiency for hydrogen also.This works regardless of the actual value of reservoir native helium concentration, which can range from a few ppm to 7% (Grynia and Griffin, 2016), as long as its concentration is monitored through time.While helium concentrations are rarely reported, a study that was specifically aimed at investigating the existence of helium in North Sea reservoirs found helium concentrations in all 9 gas reservoirs sampled (Hooker et al., 1985), including some considered in this study (Frigg, Little Dotty, Ninian, Hewett, Leman, and Viking fields).Helium concentrations have also been reported for the Amethyst (0.36%) (Garland, 1991) and Twyford (1.6%) (Gluyas et al., 2018) fields.We thus suggest that periodic sampling of reservoir gas for monitoring helium concentration evolution along with the associated noble gas and isotope analysis could be used as an improved indicator of sealing efficiency within a prospective UHS site.

Working gas ratio (𝑈 𝐺)
The recoverable gas to original gas in place ratio, not limiting, but still poses a critically important parameter for storage capacity calculations.As the exact proportion of cushion gas required will vary for different fields, a study on seasonal hydrogen storage within the Rough gas field (Amid et al., 2016) suggests a value of 0.5 is suitable for maintaining necessary reservoir pressures (Tarkowski et al., 2021).As demonstrated by Mouli-Castillo et al. (2021), fields where the ratio of recoverable gas to original gas in place exceeds 62.5% will have a working gas fraction of 0.5, and fields where the ratio is less than 62.5% will have a working gas fraction of 0.8 of the recoverable gas proportion, thereby ensuring that hydrogen accounts for at least 20% of the cushion volume (Misra et al., 1988).The working gas storage fraction UG is calculated from Eq. ( 2), where RG denotes the recoverable natural gas volume, and thus the assumed maximum volume of hydrogen that can be injected and withdrawn during the cyclic storage process, whereas OGIP, the original gas in place (Mouli-Castillo et al., 2021), is the sum of the recoverable and residual natural gas volumes. (2)

Database of UKcs gas reservoirs
The screening criteria described above are subsequently applied to 96 UKCS gas fields.The majority of these reservoirs are located in the Southern North Sea (SNS), with comparably few in the oil/condensatedominated Northern North Sea (NNS).Additionally, 6 oil fields with significant gas caps, where gas has historically been produced as a commodity, were deemed suitable for inclusion (Gluyas and Hichens, 2003).The storage capacities reported only take into consideration the gas cap volume, meaning that hydrogen storage is not considered within the oil-bearing zone.The NNS fields comprise middle Jurassic to Lower Cretaceous formations (Gluyas and Hichens, 2003), whereas those in the SNS are predominantly Triassic and Permian in age (Goffey et al., 2020).The overwhelming majority of the screened fields comprise clastic sediments, predominantly, sandstones.Although there is some evidence that carbonaceous rocks may promote hydrogeninduced dissolution (Bensing et al., 2022;Al-Yaseri et al., 2022), only two carbonaceous mineral-bearing fields were included in this analysis: the Joanne and Judy fields, composed of the Ekofisk (Waters et al., 2007) and Tor fm (formation) (Whittaker and Green, 1983), respectively.The latter also contain clays, mudstones, and siltstones in varying compositions (Waters et al., 2007;Whittaker and Green, 1983;Deegan and Scull, 1977;Waters et al., 2009), so that further geochemical assessment is required.
Table B.5 presents the field properties involved in the high-level screening and working gas fraction ( ).Additionally, the reservoir formation lithology accompanied by its average formation porosity and permeability values, the reservoir depth (to crest, m.), and initial pressure (bar) are reported.The initial pressure indicates the maximum allowable pressure and thus the maximum quantity of hydrogen that can be stored, although lower pressures may have to be applied to avoid hydrogen leakage across micro-fractures (Section 6.2.1).Moreover, the recoverable to original-gas-in-place ratio is reported ( ∶  ) which is in turn used for the calculation of the working-to-cushion gas ratio ( ).As shown in Table B.5 the average value of  ∶  is approximately 0.62 meaning that significant quantities of native gas remain in the porous media, which can be used as cushion gas.Furthermore, this  ∶  value suggests that significant quantities of natural gas have been historically produced highlighting the productivity and favourable petrophysical characteristics of these fields.The exact gas compositions of these fields are rarely reported but still are worth considering on a site-specific basis, to assess the likelihood of mixing.
Field inclusion within this study was limited by public data availability.The list of reservoirs analysed is presented in Table B.6.

Discussion
Of the 96 offshore UK hydrocarbon fields investigated, 32 were deemed suitable for UHS based on the proposed screening criteria (Table B.5). Fields lacking suitable public data availability have been excluded from this study, hence more fields than those proposed may also meet the defined criteria and prove suitable for UHS.
Fig. 3 indicates the geographic distribution of selected potential UHS sites, with respect to UK gas terminals, existing natural gas pipelines, as well as operational, planned, and under-construction offshore wind farms.More than 60% of the selected fields have existing direct pipelines to the UK grid, therefore reducing the cost of potential storage projects.Ravenspurn North, located in the southern North Sea, has the largest hydrogen storage capacity of 89.1 TWh, 50% of the total stored mixture, with residual natural gas accounting for cushion gas requirements.The reservoir comprises Leman sandstones with varying porosity and permeability, which in parts of the field is inhibiting to aqueous phase flow but not gaseous phase conductance (Anon, 1991).Roughly one-third of its reserves have been depleted, and the remaining N 2 -rich natural gas (2.5% mol) could serve as a cushion gas (Anon, 1991), although further analysis is required regarding the favourable characteristics of cushion gas types.The field with the second largest hydrogen storage capacity is the Rhum field, which lies in the NNS.In less than 20 years its two existing wells were depleted causing reservoir pressure to decline from roughly 850 bar to 470 bar so that a new well was recently drilled and is currently being produced (Anon, 2021d).This rapid and significant depletion highlights the favourable reservoir characteristics to gas flow conductance.The field's close proximity to the St. Fergus terminal, from which the produced gas is currently being transported, is also advantageous for UHS (Anon, 2021d).The third largest field is the Cygnus field, being the largest gas reservoir in the SNS with a hydrogen storage capacity of 45 TWh.Located 150 km off the coast of Lincolnshire, the main Leman sandstone formation exhibits an average permeability of 150 mD and porosity of 11% (Dredge and Source: With information from Anon (2022dAnon ( , 2021cAnon ( , 2023b)).
Marsden, 2020).Gas production started in 2016 with well production exceeding expectations leading to 150  of produced gas within two years (Catto et al., 2018) highlighting its favourable gas-flow conductance properties.Being developed by horizontal wells (Catto et al., 2018) and having a 55-km pipeline transmission system built to transport gas to the nearby Bacton terminal in Norfolk (Anon, 2019b) are advantageous traits for UHS too.Similarly, the remaining fields reflect high storage capacities (TWh order of magnitude) (Table B.5), proven gas productivity as indicated by the mean 0.62  ∶  , as well as remaining native gas that can be used as cushion gas.Overall, the five fields with the highest storage potential have a cumulative energy storage capacity of 270 TWh, which was greater than the annual natural gas requirements of UK power stations in 2021 (Anon, 2022e).While such demands are unlikely to exist for hydrogen in the near future, it highlights the significant potential of UHS for storing large amounts of energy in the form of hydrogen.
While most fields are located in the SNS, results suggest good geographic availability among suitable fields, important for reducing transportation costs and energy security.Many of the selected gas fields have existing pipeline infrastructure connecting them to gas processing terminals, which may be repurposed for hydrogen transportation in the future (Anon, 2020).Moreover, the proximity of SNS fields to existing and planned offshore wind farms (Anon, 2022d(Anon, , 2021c(Anon, , 2023b) (Fig. 3) highlights the potential economic benefits of collocating offshore hydrogen generation and storage, particularly for reducing energy lost through curtailment (Giampieri et al., 2023).One notable result is that no fields in the East Irish Sea met the screening criteria, hence other forms of local hydrogen storage may be required in the event of widespread hydrogen uptake.
The UHS site selection criteria applied, formulated on the basis of Town Gas storage, helium storage and natural gas well integrity case studies, could be further improved by considering economic barriers to development.Following this preliminary screening, site-specific studies should be applied to the gas fields to investigate the impacts of microbial and geochemical environments on hydrogen losses; as well as a thorough assessment of the reservoir caprock sealing efficiency, considering tectonic setting, micro-/fracture patterns, heterogeneity (Uliasz-Misiak et al., 2021) and containment security of native helium within the formation.In addition, individual well tests should be undertaken to determine the robustness of plugged wells and ensure that caprock integrity has not been compromised as a result of historic production.Moreover, the in-situ fluid compositions of natural gas and resident brines should be assessed in order to ascertain their impact on diffusion and hydrogen reactivity as suggested by the Beynes aquifer case study.Depending on the required loading quantities of hydrogen as well as their timing the storage capacity and dimensions of the field may factor in optimal site selection, which should be investigated on a case-by-case basis.

Conclusions
The reviewed case studies suggest the feasibility of hydrogen storage mixtures in deep geological formations but also highlight some of the key challenges for stable and efficient UHS.In this study, it was attempted to parametrise these associated challenges as site-selection screening criteria for UHS in depleted gas reservoirs.
• Caprock integrity directly affects containment security and thus, should constitute a high-level screening criterion alongside storage capacity and depth.Caprock lithology was used in this study to that end.However, as indicated by the HIEs the tectonic setting may adversely affect the sealing efficiency causing leaks of varying magnitude.Fracking and micro-fracking patterns instigated by well-drilling or hydrocarbon production should be investigated on a case-by-case basis, to both confirm sealing efficiency and ascertain the maximum allowable operating pressure.• Helium exploration analyses showed that all investigated North Sea reservoirs contained helium in varying concentrations.Thus, the existence of helium, which is currently being disregarded, could be used as an additional high-level indicator for containment security.Evidence that the smaller and more diffusive helium molecules are effectively contained within the prospect storage formation suggests hydrogen could be also.
• Microbial reactions have adversely affected town gas storage projects in the past resulting in significant hydrogen losses meaning that their consideration is of great importance.Formation temperature and salinity were selected in this study as screening criteria pertaining to microbial reaction-restriction.Further analysis is also required to determine the effect of formation lithology and cushion gas composition on the reactivity of stored hydrogen.• Well integrity was also found to be an important source of loss in former NGS projects in depleted hydrocarbon reservoirs and was incorporated into the screening criteria through the number and the age of the existing wells.Hydrogen exposure along with extreme subsurface conditions can also adversely affect wellbore integrity, although recent field-scale and lab applications indicated no impairment.Further research is required to fully understand the effect of hydrogen exposure under HPHT conditions on the individual wellbore components.• Overall, 32 offshore gas reservoirs many with existing pipeline infrastructure, across the North Sea were found to satisfy the developed criteria; providing a range of hydrogen energy capacities, with varying cushion gas requirements and proximities to existing and planned offshore wind farms.

Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Nikolaos        (Förster et al., 2006) 100 (Förster et al., 2006) Chemical and microbiological reactions.Loss of CO and gain in CO 2 as well as CH 4 and H 2 (Liebscher et al., 2016).
Similar to Reskajeage Test.
Leak across water-filled micro-fracks can be effectively reduced by reducing the maximum operating pressure.
Large no. of abandoned obsolete wells can induce H 2 leaks.

Table B.5
Screened prospect UHS sites, based on the criteria developed in this study.

Fig. 3 .
Fig. 3. Map showing offshore gas fields that meet the designated site-selection criteria for UHS along with the currently operating, planned, and under-construction offshore wind farms.Source: With information fromAnon (2022dAnon ( , 2021cAnon ( , 2023b)).

Fig. A. 5 .
Fig. A.5. Gas mixture composition fluctuations during the injection and recovery periods throughout the 16-year life period of the Ketzin Aquifer Storage site.Source: Modified from Liebscher et al. (2016).

Fig. A. 7 .
Fig. A.7. Schematic representation of the B34 borehole and the inclined fractured zone at which the helium injection took place.Source: Modified from Gascoyne and Wuschke (1997).

Table 2
Type of storage, depth, and pressure for various underground hydrogen mixture storage projects across the world.
Diamantakis reports financial support was provided by UK Industrial Decarbonisation Research and Innovation Centre.Omid Shahrokhi reports financial support was provided by UK Industrial Decarbonisation Research and Innovation Centre.Sudhagar Pitchaimuthu reports was provided by UK Industrial Decarbonisation Research and Innovation Centre.John Andresen reports financial support was provided by UK Industrial Decarbonisation Research and Innovation Centre.

Table B .4
Summary of reviewed case studies.